OpenAI released an update to Daybreak, its cybersecurity toolkit that the company first launched in June. Daybreak is designed to give cybersecurity professionals access to models, tools, and workflows aimed at defending against an increasing number of attacks that leverage artificial intelligence.
The recent update introduces a new model focused specifically on cyber defense, intended to simplify and accelerate protective tasks for researchers and incident responders. At the same time, OpenAI reorganized Daybreak into two subscription tiers: Blue and Red.
Blue and Red: different levels of access
Both tiers provide limited access to some of OpenAI's cybersecurity models; those limitations are in place to reduce the risk of misuse. The Blue tier offers more basic capabilities and is mainly suited for incident response, malware analysis, and verification of patches and remediations.
The Red tier grants broader access to OpenAI’s cyber tools and therefore could be more dangerous if misused. It includes models tuned for security testing and vulnerability research, intended for use cases that require deeper technical capabilities.
GPT-5.6-Cyber and restricted availability
Only the Red package includes the new GPT‑5.6‑Cyber model, which is built on top of GPT‑5.6 Sol. Because this language model could be potentially risky, OpenAI does not make it generally available; access is limited to the company’s “trusted partners.”
Why this matters
The update responds to the growing trend of AI-assisted cyberattacks by offering defenders purpose-built models and workflows while attempting to limit potential abuse through tiered access and vetting. TechCrunch reported the changes, and the news has been discussed in other outlets as well, highlighting the balance OpenAI is trying to strike between enabling cybersecurity work and preventing harmful use.
